For AI-first teams
Clario, right inside your editor.
The same humanize and authenticity engine that powers the app is also a Model Context Protocol server. Point Claude Code, Cursor, or any MCP client at it, and your agent can score and re-voice copy without leaving the terminal.
authenticity_score { text, voice? }Scores any text 0 to 100 for how human it reads, with the AI tells flagged span by span. Optional voice to check against.
humanize { text, voice? }Rewrites text so it reads human and, if you pass a voice, sounds like that brand. Returns the rewrite plus what changed.
Add to your MCP client
{
"mcpServers": {
"clario": {
"command": "bun",
"args": ["run", "/path/to/clario/mcp/index.ts"],
"env": { "DEEPSEEK_API_KEY": "sk-your-key" }
}
}
} - Clone the repo and set your DeepSeek key in the config above.
- Drop the snippet into your MCP client settings (Claude Code, Cursor, and others).
- Ask your agent to "score this for AI slop" or "humanize this in our brand voice."
Source lives in mcp/ in the repo. It reuses the exact prompt builders behind the web app, so the terminal and the browser give the same result.